Early exposure to infections increases the risk of allergic rhinitis—a systematic review and meta-analysis

The search strategies used the PICO principle to ensure that the retrieved journal-published literature was as comprehensive as possible. Search terms were including medical subject heading terms and text words related to subjects were developed in Pubmed and then adapted for Pubmed, Embase, Web of science, Cochrane, Sinomed, CNKI, Wanfang Database, and VIP from inception through April 30, 2022,the search terms were as follows:antenatal, prenatal, pregnancy, pregnant, perinatal, gestational, maternal, mother, newborn, infant, early life, toddler, febrile, infection, infestation, rhinitis, allergic, rhinitis, allergic, seasonal, allergic rhinitides, hayfever, hay fever. No publication, population, or language restrictions were applied, and attention was paid to checking the list of references on relevant topics. In addition, it was impossible to contact authors to have access to the full text or original data. Search terms and strategies are described in Additional file 1.

Titles and abstracts of all potentially eligible articles retrieved from each database and managed in Endnote X9 software. The literature criteria employed for the meta-analysis included the following: (1). Study population: children aged 0–18 years old. (2). Study types: cohort studies, cross-sectional studies, and case–control studies. Exposure factors: infection within 2 years after birth or maternal infection during pregnancy. (4). Outcome: clearly the offspring have AR, the relative risk (RR), hazard ratio (HR), or odds ratio (OR) and their confidence intervals can be obtained, or enough data to calculate them. The exclusion criteria were as follows: (1). The full text or original data are not available. (2). Publication languages other than Chinese or English. (3). Duplicate publications. (4). Reviews, systematic reviews or meta-analyses, conference abstracts, research protocols. (5). Viral skin infections (only one study). (6). Low-quality research.

The quality of the cohort studies and case–control studies was measured with the Newcastle–Ottawa Scale (NOS) [17], and cross-sectional studies were assessed using the Agency for Healthcare Research and Quality [18] (AHRQ). The NOS scale has a total score of 9 points, including three aspects: study population selection, comparability, and outcome, of which comparability can be scored up to 2 points, and studies with scores < 5 points are considered high risk of bias studies. The AHRQ scale has a total score of 11 points with 11 items, answering “yes” can be scored as 1 point, and answering “unclear or no” is scored as 0 points. The quality scale is divided into 0–3 points for low quality, 4–7 points for medium quality, and ≥ 8 points for high quality. In this study, the quality of literature assessed as moderate to high quality was included in the analysis (Score ≥ 6 points).

The parameters and data were extracted using a standardized spreadsheet from each study, including author, year, study country, study type, study object, sample size, exposure factors, diagnostic method, and effect values. Two researchers independently screened the literature and extracted data according to the selection of the study. After each phase of the screening and data extraction process, any disagreements regarding records between two researchers were resolved by discussion or by consulting a third investigator if consensus could not be reached.

RevMan 5.4 and Stata 16.0 software were used for statistical analysis, and the OR and 95% confidence interval (95% CI) used in the meta-analysis P ≤ 0.05 was considered to be statistically significant. Statistical heterogeneity across studies was tested by the Q statistic and I² value. If no significant heterogeneity was examined (I² < 50% and P > 0.1), pooled estimates were calculated using a fixed‐effects model; otherwise, a random‐effects model was adopted [19, 20]. Analysis of the risk relationship between different infections and AR in children. First, if a study reported different exposures;Second, if the study was stratified by exposure factors and study participants (number of infections, different periods of infection, and different ages of children) without providing overall estimates of infection and AR,the effect estimate and 95% CI from the literature were combined at first and as the final extracted effect into meta-analysis. Moreover, a subgroup analysis was performed,and sensitivity analysis was estimated by omitting every study individually. If the heterogeneity among the studies decreases after excluding one study, it shows that this study is the cause of the heterogeneity. Publication bias was assessed using funnel plots and Begg’s test of bias, with P < 0.05 indicating significant publication bias.

In total, 3866 studies were identified according to our search strategies. A total of 249 duplicate studies were excluded, and 3547 studies were inconsistent with the research purpose after reviews of the titles and abstracts. Then, 70 studies were screened and required full-text reading for detailed evaluation. Finally, 19 studies were included in this meta-analysis after full text review. More specific screening details are shown in Fig. 1 and totally 51 references were excluded and excluded reasons are shown in the Additional file 2. Among these 19 studies, 14 studies with a total sample size of 78,426 reported evidence of infection within 2 years of age and subsequent AR in children, including 7 cohort studies [15, 21‐26], 5 cross-sectional studies [27‐31] and 2 case–control studies [16, 32]. Five studies with a total sample size of 82,256 reported evidence of infection during pregnancy and AR in offspring, including 3 cohort studies [33‐35], 1 cross-sectional study [36] and 1 case–control study [37]. Except for the study of Thomson 2010 [21], the quality evaluation of all the included studies was of medium to high quality. This study used follow-up parental reports for infection exposure and AR outcome determinations. The following analysis did not include this evidence because the assessments were unreliable and the sample size was not large enough. Three of the remaining 18 studies extracted crude effect sizes [22, 27, 33], one was obtained by calculation [27], and all others extracted adjusted effect sizes. Tables 1, 2 and 3 summarize the specific information of the included studies and quality evaluation, respectively.

A total of 5 studies identified the relationship between maternal infection during pregnancy and the risk of AR in offspring. Maternal infection during pregnancy was associated with an increased risk of AR in offspring (OR = 1.34, 95% CI: 1.08–1.67) in all 5 studies as a forest plots (Fig. 2). In performing subgroup analyses, infection during pregnancy was associated with a significantly increased risk of AR in offspring in 3 cohort studies (OR = 1.14, 95% CI: 1.10–1.18) and in the literature that AR was diagnosed by a doctor (OR = 1.38, 95% CI: 1.06–1.81), but there was a trend in 2 case–control studies (OR = 2.37, 95% CI: 1.00–5.61) and in the literature that AR was diagnosed by self-reporting (OR = 1.38, 95% CI: 1.06–1.81). The analysis results are as follows (Table 4).

These overall results showed that early infection within 2 years of age was closely associated with childhood AR (OR = 1.25, 95% CI: 1.12–1.40), especially in cohort studies (OR = 1.17, 95% CI: 1.06–1.29) and noncohort studies (OR = 1.51, 95% CI: 1.13–2.02). Compared with 3 papers that confirmed AR by self-reporting, the results showed that early infection within 2 years of age was significantly related to AR in outcome evaluated by ISAAC (OR = 1.17, 95% CI: 1.06–1.30) and in trend related to AR in outcome evaluated by doctor's diagnosis (OR = 2.05, 95% CI: 0.96 -4.35) (Fig. 3).

Early upper respiratory tract infection, bronchitis, lower respiratory tract infection, ear infection, gastrointestinal infection and neonatal urinary tract infection were reported in the included studies. Subtype of infection analysis showed that a history of upper respiratory tract infection was associated with an increased risk of AR in children (OR = 1.32, 95% CI: 1.06–1.65). However, neither a history of bronchitis nor lower respiratory tract infection was related to AR in children. Ear infection was associated with an increased risk of AR in children in 4 cohort studies (OR = 1.13, 95% CI: 1.04–1.22) but not in the case–control study (OR = 0.96, 95% CI: 0.87–1.06). Gastrointestinal infection increased the risk of AR in children in the cohort study (OR = 1.20, 95% CI: 1.05–1.37), but there was an obvious tendency in 3 noncohort studies (OR = 1.70, 95% CI: 0.95–3.06, p = 0.08). Urinary tract infection was not associated with an increased risk of AR in children in any type of study. The specific results are shown in Table 5.

For maternal infection, the overall sensitivity analysis was carried out by removing the literature one by one, and it was found that the effects on the combined results were stable and reliable. However, when Liao 2015 [37] was removed, the remaining heterogeneity between studies became no longer significant (I² = 0%, P = 0.39), and the results still showed that infection during pregnancy could increase the risk of developing AR in offspring (OR = 1.14, 95% CI: 1.11–1.18). Funnel plots did not show the possibility of publication bias, and Begg’s test determined this outcome (P = 0.451, P > 0.05) (Fig. 4). Publication bias test was not performed in this analysis because fewer than 10 studies were included [38].

For another infection in the first two years of life, based on the results of the above preliminary analysis, the sensitivity analysis was carried out by using the method of eliminating literature one by one. It can be seen that unstable results for the upper respiratory tract infection and greater heterogeneity between studies, when De 2005 [28] and Kang 2019 [32] studies were excluded, a significant decrease of the test for heterogeneity was observed (I² = 0%,P = 0.02), the results of all remaining studies still show that upper respiratory tract infection was associated with increased risk of AR in children (OR = 1.11,95%CI:1.02–1.22). Similarly, for the gastrointestinal tract, after excluding the study of Strachan 1996 [27], the results of the remaining three studies were (OR = 1.59, 95% CI: 1.06–2.37). The sensitivity analysis results of other factors are stable and reliable. As before, no evidence of a significant publication bias was found, which was confirmed by the Begg’s test (P = 0.246, P > 0.05). However, asymmetry can be found in funnel plots (Fig. 4). In order to solve this problem, we estimated a sensitivity analysis using the trim and fill method (adding 2 new virtual studies) showed that the results was reliable (OR = 1.15,95%CI:1.01–1.30). Thus, we found that the outcome was not affected by publication bias. (Fig. 5).